Estimation of Key PCC, Base, Subbase, and Pavement Engineering Properties From Routine Tests and Physical Characteristics

PDF Version (4.44 MB)

PDF files can be viewed with the Acrobat® Reader®

CHAPTER 4. DATA ASSEMBLY AND MODELING CONSIDERATIONS

1. Identify and assemble all relevant data from the LTPP database.
2. Evaluate the quality of the assembled data by reviewing the assembled data for inconsistencies and possible errors, while also identifying missing/suspect data items.
3. Develop methods and procedures for estimating important missing data elements and clean data by resolving anomalies and outliers in a consistent manner.
4. Select the appropriate prediction model form and identify variables that emerge as significant variables.
5. Evaluate the reasonableness of the model formulated and verify whether the predicted correlations are meaningful from an engineering standpoint.
6. Ascertain statistical correctness and suggest tentative prediction models.
7. Perform sensitivity analysis to validate the tentative models. If validation is not satisfactory, revise the model in step 4.
8. Confirm final prediction model(s).

This approach has been used successfully in previous studies and has resulted in practical prediction models. Note that steps 4 through 8 are an iterative process; therefore, the data extraction and assembly process should allow multiple revisions to the selected variables and model forms.

Data Selection for Current Study

Data in the LTPP program exist for two complementary experiments, General Pavement Studies (GPS) and SPS. GPS experiments usually exist in service pavements incorporated into the LTPP program, while the SPS experiments are mostly newly constructed or rehabilitated pavements or pavements subjected to various maintenance activities at the time they were added to the LTPP program.

The GPS and SPS experiments are as follows:

• GPS-1: Asphalt concrete (AC) on granular base.
• GPS-2: AC on bound base.
• GPS-3: Jointed plain concrete.
• GPS-4: Jointed reinforced concrete.
• GPS-5: Continuously reinforced concrete.
• GPS-6A: Existing AC overlay on AC pavements.
• GPS-6B: New AC overlay on AC pavements.
• GPS-7A: Existing AC overlay on PCC pavements.
• GPS-7B: New AC overlay on PCC pavements.
• GPS-9: Unbonded PCC overlays on PCC pavements.
• SPS-1: Strategic study of structural factors for flexible pavements.
• SPS-2: Strategic study of structural factors for rigid pavements.
• SPS-3: Preventive maintenance effectiveness of flexible pavements.
• SPS-4: Preventive maintenance effectiveness of rigid pavements.
• SPS-5: Rehabilitation of AC pavements.
• SPS-6: Rehabilitation of jointed PCC pavements.
• SPS-7: Bonded PCC overlays on concrete pavements.
• SPS-8: Study of environmental effects in the absence of heavy loads.
• SPS-9A: Validation of Strategic Highway Research Program (SHRP) asphalt specification and mix design.

Each GPS test site consists of a single 500-ft test section over which all factors remain constant. SPS-1 and SPS-2 sites usually consist of a series of adjacent 500-ft test sections with different design and material characteristics or maintenance treatments and rehabilitation strategies. The test section layouts and the material data formats in the various LTPP data tables are reported elsewhere.⁽¹⁴¹⁾

The GPS and SPS experimental sections have different data availability. Relative to SPS sections, GPS sections are older pavement projects and were not intentionally built for collecting data under the LTPP program. SPS sections, in contrast, offer more detailed information on materials and construction. Further, the materials used in GPS sections are not necessarily representative of current materials, as several elements of the material manufacturing process and material specifications have changed. With the likelihood that GPS and SPS section data could produce different models, the data extraction and data assembly processes were tailored to analyze these data in two separate groups. Data for the various models by material type were taken from the GPS and SPS sections as shown in table 9.

Table 9. LTPP sections selected to review data for each category.

The following additional points offer reasons for the selection of materials data collected for PCC models:

• GPS-1, GPS-2, and GPS-6 projects were not included because they do not include PCC layers.
• SPS-1, SPS-3, SPS-5, and SPS-9 projects were not included because they do not include PCC layers.
• SPS-4 projects contain older rigid pavements with preventive maintenance, and it was found that the data relative to index properties were less accurate than for the newer SPS projects. These data were included with the GPS experiment data because the underlying pavement data were recovered from the original GPS projects.
• SPS-6 projects consist of underlying PCC layers that have been rehabilitated and sometimes cracked or rubblized. The data for the PCC layers are in the original inventory database. As with the SPS-4 underlying PCC slab data, these data were included in the GPS data tables assembled for this study.
• Data for the overlay only were included from the SPS-7 projects because the underlying slab information is relatively less comprehensive.

Table 10 shows the LTPP data tables that were queried to obtain the data necessary for developing the prediction models. The table also lists the data elements that were obtained from each table.

Table 10. LTPP data tables queried to obtain data for review and future analyses.

Data Review

The data review process sought to verify data availability for the predictive models based on the dependent and independent variables identified in chapters 2 and 3 of this report. The assembled data were reviewed to identify anomalies and missing data elements. The review evaluated whether the following occurred:

• All required data were available.
• Available data were reasonable (i.e., trends and values of the data are as expected according to literature and engineering knowledge).
• Missing data items and the number of missing datasets significantly impaired the ability to develop required models for this study.
• Missing data items could be computed, assumed, or obtained from other readily available sources.

To enable such an evaluation, data were assembled for each 500-ft section in the database. As explained earlier, the data for each GPS section correspond to the test site. However, for SPS sections, it was reasonable to use the index properties of a given material from each site to correlate the data to the material properties from multiple test sections that were constructed using the same material. The treatment of PCC data from SPS-2 sections are discussed in detail in the PCC model development section in chapter 5 of this report.

Also included in developing the reference system were layer number, layer type, and construction number.

Key Findings From Data Review

A detailed unpublished report summarizing the results of the data review process was completed in phase I of this study and submitted to FHWA. The following conclusions were presented:

• All key data elements required for model development according to the literature were available.
• For PCC materials, SPS sections have more extensive data than GPS sections, as expected. This necessitates different model considerations and the inclusions of different parameters in the models using these two datasets.
• While independent variables exist for most predictive models, there are no dependent variable values in the database for parameters that are representative of the construction or site conditions. These parameters include PCC zero-stress temperature, EI for JPCP design, erosion in CRCP design, base friction coefficient for CRCP design, and deltaT for JPCP and CRCP designs. These parameters were never intended for inclusion in the LTPP database and are not typically measured in the field during construction. These variables are specific to MEPDG inputs and were estimated through correlations during the development of the MEPDG. It was therefore necessary to use MEPDG calibration data to verify or refine these models.
• Most stabilized materials are referred to as treated bases in the LTPP database, and only compressive strength data exist for these materials. Index properties were not included in the database. Predictive models therefore were not developed for a majority of the stabilized base materials. However, the following materials data for LCB layers in SPS-2 sections are more comprehensive:
  • LTPP Standard Data Release 23.0, which was evaluated during phase II of this project, contained no elastic modulus data for LCB materials.⁽³⁾
  • LTPP Standard Data Release 24.0, released after this study was initiated, contained elastic modulus data for LCB layers in SPS-2 sections.⁽¹⁴²⁾
• LTPP Standard Data Release 23.0 contains data beyond that utilized in a previous LTPP study investigating various models and predictive variables to estimate resilient modulus of unbound materials.^(3,135)

All data available were of adequate quality and completeness for use in model development. However, some missing data and data anomalies were also identified. The specific issues encountered and the methods used to overcome them is discussed in chapter 5 for each material type and material property. A summary of data availability is presented in table 11 for each material type and in table 12 for models to predict rigid pavement design features.

Table 11. Summary of data availability for material property predictive models.

N/A = Not applicable; LS = Low strength; HS = High strength.

*SPS data have been summarized by experiment type. SPS-2-LS and SPS-2-HS refer to SPS-2 sections that have been built with low-strength and high-strength concrete mixes.

**Stabilized material data contain only compressive strength data and elastic modulus data for LCB materials.

Table 12. Summary of MEPDG calibration sections to develop predictive models for rigid pavement design features.